I implemented a proof of concept Rubinius-style backtrace, but recently removed it as it was only half done. See b54ca29.

We should be able to precisely match the Rubinius backtrace format as in Truffle we have a very good backtrace. Our MRI style backtrace is designed to be character-for-character identical, and we'd like to achieve the same for Rubinius with the obvious exception that we won't include the source of their core methods. Colour would be good as well.

We could then add an option to include the value of locals in each activation, as we do in our debug formatter. See 03c0ca6.